CompareToolsMethod @niklaslenz_ai
Home / Text
Text

Text models, all memory classes

Every complete run of a text model, grouped by the memory it needs. Pick a memory class to narrow it down; the default is all of them. Click a number for the raw output, a row for the recipe.

Memory class
Type
16 GB
The 16 GB page →

RTX 5060 Ti 16 GB · 1 run

#Creator Setup Overallof 100 Agent testbrowser check Speedtok/s Qualitymean of 7 tests Memorylargest context
1Mia's AI LabQwen3.8-27B1× RTX 5060 Ti77.5 of 100rank 1 of 1as of 20 Oct 2026Page built in the agent test7 / 1052.9≈ 13.5 s for 500 wordscode 53.6 · prose 52.166.1%mean of 7 tests115Ktokens tested

Scores are relative to the best run of this class on 20 Oct 2026. Compare runs side by side.

32 GB
The 32 GB page →

RTX 5090 32 GB · 9 runs

#Creator Setup Overallof 100 Agent testbrowser check Speedtok/s Qualitymean of 7 tests Memorylargest context
1nerouedQwen3.8-27B1× RTX 509073.5 of 100rank 1 of 9as of 25 Oct 2026Page built in the agent test7 / 10159.6≈ 4.7 s for 500 wordscode 169.0 · prose 150.265.3%mean of 7 tests240Ktokens tested
2Mia's AI LabQwen3.8-27B1× RTX 509066.7 of 100rank 2 of 9as of 25 Oct 2026Page built in the agent test7.5 / 10149.3≈ 4.8 s for 500 wordscode 154.5 · prose 144.267.6%mean of 7 tests128Ktokens tested
3BobClawblawQwen3.8-27Bnew1× RTX 509063.4 of 100rank 3 of 9as of 25 Oct 2026Page built in the agent test7.5 / 10253.4≈ 2.9 s for 500 wordscode 268.2 · prose 238.762.9%mean of 7 tests256Ktokens tested
4kacper-daftcode (Qwentin)Qwen3.8-27Bnew1× RTX 509059.6 of 100rank 4 of 9as of 25 Oct 2026Page built in the agent test8 / 10148.7≈ 4.9 s for 500 wordscode 158.3 · prose 139.267.9%mean of 7 tests128Ktokens tested
5Peasant Smith (Project Maya)DeepSeek-V4.1-Flash-10171× RTX 509052.4 of 100rank 5 of 9as of 25 Oct 2026Page built in the agent test9 / 1020.9≈ 35.1 s for 500 wordscode 22.9 · prose 18.968.0%mean of 7 tests128Ktokens tested
6The Kaitchup (Benjamin Marie)Qwen3.8-27B-SLQ-EAR98new1× RTX 509051.4 of 100rank 6 of 9as of 25 Oct 2026Page built in the agent test8 / 1067.2≈ 10.7 s for 500 wordscode 67.2 · prose 67.164.3%mean of 7 tests73Ktokens tested
7Mia's AI LabQwen3.8-27Bnew2× RTX 5060 Ti49.0 of 100rank 7 of 9as of 25 Oct 2026Page built in the agent test8 / 1070.8≈ 10.2 s for 500 wordscode 73.1 · prose 68.567.2%mean of 7 tests256Ktokens tested
8RecursiveFlint 40B1× RTX 509044.6 of 100rank 8 of 9as of 25 Oct 2026Page built in the agent test9.5 / 10224.9code 224.8 · prose 224.938.6%mean of 7 tests256Ktokens tested
9The Kaitchup (Benjamin Marie)Qwen3.8-27Bnew2× RTX 5060 Ti44.4 of 100rank 9 of 9as of 25 Oct 2026Page built in the agent test8 / 1033.0≈ 22.2 s for 500 wordscode 33.0 · prose 33.065.6%mean of 7 tests92Ktokens tested

Scores are relative to the best run of this class on 25 Oct 2026. Compare runs side by side.

128 GB
The 128 GB page →

1× DGX Spark (GB10) · 8 runs

#Creator Setup Overallof 100 Agent testbrowser check Speedtok/s Qualitymean of 7 tests Memorylargest context
1Mia's AI LabQwen3.8-Flash-Next1× DGX Spark76.3 of 100rank 1 of 8as of 25 Oct 2026Page built in the agent test8.5 / 1042.8≈ 16.9 s for 500 wordscode 43.2 · prose 42.567.1%mean of 7 tests256Ktokens tested
20xWhiteMageThinkingCap-Qwen3.8-27B-Kearuga1× DGX Spark73.3 of 100rank 2 of 8as of 25 Oct 2026Page built in the agent test8 / 1048.3≈ 15.6 s for 500 wordscode 54.0 · prose 42.564.4%mean of 7 tests256Ktokens tested
30xBakeerOrnith-1.5-122B1× DGX Spark69.6 of 100rank 3 of 8as of 25 Oct 2026Page built in the agent test8 / 1044.2≈ 15.5 s for 500 wordscode 43.9 · prose 44.670.8%mean of 7 tests256Ktokens tested
4Mia's AI LabOrnith-1.5-122B1× DGX Spark67.4 of 100rank 4 of 8as of 25 Oct 2026Page built in the agent test8 / 1042.3≈ 16.9 s for 500 wordscode 43.5 · prose 41.169.0%mean of 7 tests256Ktokens tested
50xWhiteMageThinkingCap-Qwen3.8-27B-Kearuga1× DGX Spark65.1 of 100rank 5 of 8as of 25 Oct 2026Page built in the agent test9 / 1045.5≈ 15.6 s for 500 wordscode 45.7 · prose 45.465.5%mean of 7 tests256Ktokens tested
6ViC305 (Victor Cruz)Qwen3.8-Flash-Nextnew1× DGX Spark57.9 of 100rank 6 of 8as of 25 Oct 2026Page built in the agent test8.5 / 1056.2≈ 12.6 s for 500 wordscode 58.4 · prose 54.066.2%mean of 7 tests128Ktokens tested
7plotarmordevNemotron 3.5 Super VL 120Bnew1× DGX Spark56.0 of 100rank 7 of 8as of 25 Oct 2026Page built in the agent test8 / 1041.5≈ 16.7 s for 500 wordscode 40.6 · prose 42.457.5%mean of 7 tests128Ktokens tested
8BottleCap AIThinkingCap-Qwen3.8-Flash-Nextnew1× DGX Spark56.0 of 100rank 8 of 8as of 25 Oct 2026Page built in the agent test8.5 / 1030.5≈ 24.3 s for 500 wordscode 30.6 · prose 30.367.4%mean of 7 tests256Ktokens tested

Scores are relative to the best run of this class on 25 Oct 2026. Compare runs side by side.

256 GB
The 256 GB page →

2× DGX Spark (GB10) · 4 runs

#Creator Setup Overallof 100 Agent testbrowser check Speedtok/s Qualitymean of 7 tests Memorylargest context
1Mia's AI LabDeepSeek-V4.1-Flash-10172× DGX Spark86.8 of 100rank 1 of 4as of 24 Oct 2026Page built in the agent test8.5 / 1068.5≈ 10.9 s for 500 wordscode 72.0 · prose 65.068.2%mean of 7 tests128Ktokens tested
2bertholomusDeepSeek-V4.1-Flash-10172× DGX Spark82.2 of 100rank 2 of 4as of 24 Oct 2026Page built in the agent test8 / 1070.6≈ 10.0 s for 500 wordscode 70.5 · prose 70.666.1%mean of 7 tests128Ktokens tested
3sudoingXDeepSeek-V4.1-Flash-10172× DGX Spark78.7 of 100rank 3 of 4as of 24 Oct 2026Page built in the agent test8.5 / 1063.6≈ 11.6 s for 500 wordscode 64.9 · prose 62.269.1%mean of 7 tests128Ktokens tested
4Mia's AI LabGLM-5.3-Flash2× DGX Spark73.3 of 100rank 4 of 4as of 24 Oct 2026Page built in the agent test9 / 1049.9≈ 14.0 s for 500 wordscode 51.3 · prose 48.662.4%mean of 7 tests128Ktokens tested

Scores are relative to the best run of this class on 24 Oct 2026. Compare runs side by side.

256 GB Mac
The 256 GB Mac page →

Mac Studio M5 Ultra 256 GB · 3 runs

#Creator Setup Overallof 100 Agent testbrowser check Speedtok/s Qualitymean of 7 tests Memorylargest context
1oMLX (Jun Kim)Qwen3.8-Flash-NextMac Studio M5 Ultra76.5 of 100rank 1 of 3as of 23 Oct 2026Page built in the agent test8.5 / 10152.5≈ 5.4 s for 500 wordscode 175.8 · prose 129.164.8%mean of 7 tests256Ktokens tested
2gnahzhangDeepSeek-V4.1-Flash-10171× Mac Studio M5 Ultra61.2 of 100rank 2 of 3as of 23 Oct 2026Page built in the agent test8 / 1063.8≈ 11.2 s for 500 wordscode 67.7 · prose 60.060.0%mean of 7 tests128Ktokens tested
3antirez (Salvatore Sanfilippo)DeepSeek-V4.1-Flash-10171× Mac Studio M5 Ultra55.6 of 100rank 3 of 3as of 23 Oct 2026Page built in the agent test8.5 / 1054.5≈ 13.2 s for 500 wordscode 54.5 · prose 54.569.7%mean of 7 tests128Ktokens tested

Scores are relative to the best run of this class on 23 Oct 2026. Compare runs side by side.

512 GB
The 512 GB page →

4× DGX Spark (GB10) · 2 runs

#Creator Setup Overallof 100 Agent testbrowser check Speedtok/s Qualitymean of 7 tests Memorylargest context
1bertholomusDeepSeek-V4.1-Flash-10174× DGX Spark90.7 of 100rank 1 of 2as of 22 Oct 2026Page built in the agent test9 / 1081.8≈ 9.1 s for 500 wordscode 82.8 · prose 80.866.6%mean of 7 tests128Ktokens tested
2Mia's AI LabDeepSeek-V4.1-Flash-10174× DGX Spark79.4 of 100rank 2 of 2as of 22 Oct 2026Page built in the agent test9 / 1051.2≈ 13.6 s for 500 wordscode 50.4 · prose 52.066.7%mean of 7 tests128Ktokens tested

Scores are relative to the best run of this class on 22 Oct 2026. Compare runs side by side.

Compare up to three runs →